Background on Florida Gulf Bay Scallops

Most Florida Gulf bay scallops live less than 2 y and spawn only one reproductive season (Loosanoff & Davis 1963, Sastry 1965, Castagna & Duggins 1971, Barber & Blake 1983). Bay scallops are hermaphroditic. Each individual spawns eggs and sperm in separate pulses (Arnold et al. 2005a, Arnold et al. 2005b), and self-fertilization can result in offspring production (Wilbur 1995). Florida Gulf bay scallops aggregate during the spawning season (Marelli et al. 1999) and aggregations spawn more or less simultaneously (Levitan & Petersen 1995, Arnold et al. 1998a, Marelli et al. 1999). Gametes are viable for only minutes to hours (Levitan 1995). Bay scallops have threshold densities below which spawning and recruitment do not support annual population renewal (Allee effect) (Greenawalt-Boswell et al. 2007).

The principal spawning season extends from September or October through December or January (Geiger et al. 2010). Peak spawning is most common in October, but can occur during any of those months or may be spread over several months (Arnold et al. 1997). Secondary local spawning bursts may occur in spring or early summer, and very low levels of spawning may occur during other months, but these occurrences yield little successful spat recruitment (Arnold et al. 1998a, Arnold et al. 2009, Geiger et al. 2010; Florida Fish and Wildlife Conservation Commission, Fish and Wildlife Research Institute (FWRI), unpubl. data). Because spawning season is protracted and duration of the larval phase is 6-14 days (Sastry 1961, Sastry 1965, Arnold et al. 2005b), spawning and recruitment seasons overlap.

Bay scallop preference for Thalassia testudinum König/ Syringodium filiforme Kutz (hereafter, seagrass) beds as benthic habitat is strong. Presettlement larvae must attach to seagrass blades to metamorphose into spat (Thayer & Stuart 1974, Barber & Blake 1983, Pohle et al. 1991), which remain attached to the blades until about 20-mm in shell height (Geiger et al. 2010). Juveniles and adults move about but travel only short distances (Barber & Blake 1983), remaining within seagrass meadows for the duration of their lives. High rainfall causes flooding that kills bay scallops by reducing salinity and increasing turbidity to intolerable levels (Tettelbach & Rhodes 1981, Leverone 1993); eggs and larvae are more sensitive to these environmental conditions than are juveniles and adults (Tettelbach & Rhodes 1981). Toxic dinoflagellate outbreaks (blooms) such as red tide kill bay scallops in any life phase (Marko & Barr 2007) through ingestion (Peterson & Summerson 1992) or by rendering large areas of water near seafloors hypoxic for extended time periods (Landsberg 2002). Bay scallop abundances can take years to recover from severe blooms (Peterson & Summerson 1992, Leverone et al. 2010). In the Florida Gulf, Karenia brevis forms blooms essentially annually (Flaherty & Landsberg 2011).

Bay scallops are distributed patchily in the Florida Gulf (Marelli & Arnold 2001, Arnold et al. 2005a). Fluctuations in abundance have long characterized Florida Gulf bay scallop subpopulations (Arnold et al. 2005a); but, many historically abundant subpopulations, particularly those from Tampa Bay southward (Fig. 1), have declined enduringly in abundance and have become at least intermittently extinct (Arnold et al. 2005a, Leverone et al. 2006) as a result of habitat degradation (Leverone et al. 2006), red tide blooms (Arnold et al. 1998b), and overfishing (Marelli & Arnold 2001). Statewide, abundance had generally declined to a tiny fraction of its original size by the late 1980s (Geiger et al. 2006). By 1994, only the Deadman Bay and St. Joseph Bay (Fig. 1) subpopulations attained healthy abundance levels (≥25 scallops/600 m² (Arnold et al. 2005a, Greenawalt-Boswell et al. 2007)), but only during some years (Arnold et al. 1998a).

Florida Gulf bay scallops supported robust commercial and recreational fisheries during the 1950s (Murdock 1955, Rosen 1959). The abundance declines prompted the initiation of fishery management in 1985. The commercial fishery was closed statewide in 1994. In 1995, all recreational harvest was eliminated from the Suwannee River southward and from the mouth of St. Joseph Bay westward (Fig. 1) (Arnold et al. 2005a, Geiger et al. 2006). Between the Suwannee River and St. Joseph Bay, recreational harvest was permitted but restricted (very short harvest season, end of July through beginning of August; daily limits on amount taken per boat, but no limit on the number of boats that could be deployed (Arnold et al. 1998a)).

Before 1998, attempts to supplement Florida Gulf bay scallop stocks were limited, sporadic (Arnold et al. 1999, Arnold 2001, Arnold et al. 2005a, Leverone et al. 2006, Arnold 2008, Leverone et al. 2010), and focused principally on Tampa Bay (Lu & Blake 1997) (Fig. 1). Low abundances after the recreational harvests from 1994 through 1996 prompted initiation of the first major stock enhancement effort during fall 1998—the stocking of spawner restoration stocks in Homosassa Bay, Anclote Estuary, and Tampa Bay (Arnold et al. 2005a, Wilbur et al. 2005). The first recruits from that stocking would have appeared in 1999.

Based on distributional patterns, prehistory and contemporary morphological variation, and spat recruitment patterns of Florida Gulf bay scallops, Arnold and colleagues (Marelli et al. 1997a, Arnold et al. 1998a, Marelli & Arnold 2001, Arnold et al. 2005a, Arnold 2008) postulated that Florida Gulf bay scallops were distributed as a series of disjunct subpopulations that together formed a metapopulation. Arnold et al. (1998a) postulated that the stable, high-density Steinhatchee subpopulation (Fig. 1) might be the main provider of migrants to other subpopulations (i.e., a source subpopulation, or source). Bert et al. (2011) showed that the Florida Gulf bay scallop population genetic structure differed from that of western Atlantic Ocean bay scallops and that bay scallops from Florida Bay (Fig. 1) were Argopecten irradians ssp. concentricus and not Argopecten irradians ssp. taylorae. They also described in detail previous population genetics studies that included samples of Florida Gulf bay scallops (Blake & Graves 1995, Wilbur 1995, Marelli et al. 1997a, Marelli et al. 1997b, Bologna et al. 2001, Hemond & Wilbur 2011), including a genetic assessment of stock enhancement success (Wilbur et al. 2005). Based on comparisons of the relatively genetically distant Atlantic bay scallop (Argopecten irradians ssp. irradians) with only a few Florida Gulf bay scallop collections, those researchers concluded that Florida Gulf bay scallops, except those in Florida Bay, formed a single, more or less panmictic population. Our study was conducted to elucidate more fully the population genetic structure of Florida Gulf bay scallops and to test genetically the idea of Arnold and colleagues that those bay scallops formed a metapopulation.

Figure 1.

Bay scallop collecting locations. Large inset, location of Florida in eastern United States. Small insets, details of sampling locations and corresponding collection abbreviations (see Table 1; no inset for Hernando (HE)); hatched areas, approximate extent of sampled seagrass beds. Asterisks, locations of labeled features; x in FB inset, approximate location of Rabbit Key Basin.

The Study Area

The Florida Gulf includes waters off the south-facing coast of the Florida panhandle (hereafter, panhandle) and the west-facing coast of the peninsula, as well as Florida Bay (Fig. 1). The seafloor off peninsular Florida slowly slopes down to 200 m (1 m depth/1–2.5 km distance (Li & Weisberg 1999)). The expansive shelf buffers heavy wave action in shallow water; only intense winter cold fronts and summer tropical cyclones result in waves more than 2 m high nearshore. The nearshore seafloor is a highly fragmented patchwork of seagrass flats, low rock outcrops, and bare sand. Interspersed along the coast are bays and rivers with restricted exchange with open water; large embayments, many times behind barrier islands, with freshwater marshes or estuarine mangrove swamps grading into seagrass flats or mixed seagrass/rock/sand; and sandy beaches facing open water. West of Cape San Blas (CSB), the seafloor slopes more rapidly and wave action is stronger. Relatively quiet waters are restricted to bays—which have a patchy seafloor of interspersed mud, sand, rocky outcrop, and seagrasses—behind barrier islands with few channels to open water. Florida Bay is composed of numerous shallow basins (most <1–3 m deep) connected by sparsely distributed channels through a network of sand bars, calcareous mud/sand banks, and small islands that, together, separate the basins. The network reduces tidal range, currents, and extreme wave action. Adjacent basins can have very different benthic habitats and ecosystems.

Field and Laboratory

Between 1995 and 1998, FWRI staff collected bay scallops during scuba diving surveys conducted in June of each year, before the annual harvest season. They collected the first ±50 individuals encountered; or, if density was low, they searched or trawled, as time allowed, attempting to obtain at least 15 individuals. Multiple annual collections were made during the 4 y at some locations (hereafter, 2-y, 3-y, or 4-y subpopulations). Collections were not made at all locations in all years principally because bay scallops could not be found. Scallop samples were returned to the laboratory where, from each individual, a sample of adductor muscle, gill, and digestive gland was excised. The samples were wrapped, frozen immediately in liquid nitrogen, and stored at -80°C.

For allozyme electrophoresis, small pieces of the three tissue types were combined and homogenized in 0.1 M Tris-EDTA, pH 7.0; the supernatant was used as the enzyme source. Horizontal starch gel electrophoresis was conducted according to Selander et al. (1971). The 4 buffer systems and the staining procedures used to resolve and visualize the 18 loci assayed are listed in Bert et al. (2011, Table 2). All gels were scored by at least 2 researchers. Alleles were identified by their mobility relative to that of the most common allele (designated as 100); the numerical code was translated into an alpha code (e.g., 100 - A) for some statistical analyses.

For the mtDNA analysis, various constraints required that we omit samples from 3 low-density subpopulations (CI, CK, HE). The mtDNA analytical techniques and solution formulas used are detailed in Seyoum et al. (2003) and Bert et al. (2011). Briefly, we assayed the restriction fragment patterns produced by digesting an 833-bp fragment that included a portion of the mtDNA 12s ribosomal subunit and the NADH dehydrogenase 1 coding region with 9 restriction enzymes: Alu I, Ban II, Bgl II, BsiHKA I, HinF I, Rsa I, ScrF I, Tsp 509 I, and Taq ^α I. Entire digests were loaded onto low-melting point agarose gels and were electrophoresed. Fragment patterns were visualized by ethidium bromide staining and were photographed under UV light. Fragment sizes were determined from migration distances relative to known standards.

Statistical Analysis

Many statistical analyses were performed at both the collection level (allozyme, n = 31; mtDNA, n = 21) and the subpopulation level (all collections combined from a single location: allozyme, n = 12; mtDNA, n = 9; Table 1). Throughout this article, sample designations for collection-level analyses include both the location and year of collection (e.g., SA95) and designations for subpopulation-level analyses include only the location abbreviation (e.g., SA). When only 1 sample was collected from a location, that sample served in both collection-level (e.g., FB98) and population-level (e.g., FB) analyses.

Despite the possibility of overcorrecting for the possibility of type I statistical errors (Moran 2003, Nakagawa 2004), unless otherwise noted, significance levels of multiple tests of a single hypothesis were adjusted using the sequential Bonferroni technique (Rice 1989), which is designed to avoid type I errors. When many tests of a single hypothesis were conducted, we tempered the possibility of overcorrection for spurious significance by adjusting a levels using the number of significant tests rather than the total number of tests for the sequential Bonferroni adjustment, set significance at a higher level, or simply reported all significance levels (i.e., used a commonsense approach (Cabin & Mitchell 2000)).

Allozyme Loci

We originally used BIOSYS-1 (Swofford & Selander 1981) to calculate allele and genotype frequencies. We used GENEPOP (version 4.2 (Rousset 2008)) to calculate collection- and subpopulation-level statistics for all loci. All statistical tests were conducted on loci only in which the frequency of the most common allele was ≤0.99. To examine the spatial and temporal nature of genetic diversity, we tested each locus for homogeneity of allele and genotype frequencies between multiple collections from each location, and for homogeneity of allele frequencies among subpopulations using the exact probability test. For every collection, we compared observed genotype frequencies for each locus with Hardy-Weinberg (H-W) equilibrium genotype frequency expectations using the Markov chain method. We also compared the percentage of loci deviating from H-W equilibrium among collections, subpopulations, and years (pooling collections within years) using the R × C G-test followed by the Simultaneous Test Procedure for frequencies (hereafter, collectively, the R × C test; BIOMstat, version 3 (Sokal & Rohlf 1995).

TABLE 1.

Summary information for Florida Gulf bay scallop samples used to estimate population genetic structure.

We calculated average direct-count heterozygosity per locus (), the percentage of polymorphic loci at the P₉₅ and P₉₉ levels (frequency of most common allele ≤0.95 or 0.99, respectively), and mean number of alleles per locus () at both the collection and subpopulation levels, and tested for correlations between those variables and sample size at both levels. We examined geographical and temporal variation in these genetic diversity measures by testing for correlations between the measures and latitude, by examining the values for collections within locations for directional pattern shifts among years when possible, and by testing for significant differences among collections taken in different years and among all collections and subpopulations. For all these tests, we used the Kruskal-Wallis test to establish significance followed by the Ryan-Einot-Gabriel-Welsch Multiple Range test (hereafter, collectively, the K-W test) to locate significant differences (Statistical Analysis System, version 9.1; SAS Institute Inc.).

We also used GENEPOP to test F_IS (inbreeding coefficient) values for significant differences among collections taken in different years, among all collections, and among subpopulations composed of two or more collections. We examined collections for the possibility of genetic bottlenecking using the BOTTLENECK program (Cornuet & Luikart 1996).

We investigated temporal and spatial population genetic associations by first testing allele frequencies collectively over all loci for homogeneity between each subpopulation pair (exact probability test). We calculated pairwise Nei's genetic distances (Nei's D (Nei 1972)) and pairwise Θ;_ST values (Wright's F_ST analogue (Weir & Cockerham 1984)) over all loci (according to Slatkin (1993), in GENEPOP) between collections and between subpopulations. We used subpopulation Θ;_ST values to generate a global estimate of Θ;_ST, to estimate the average gene flow according to Wright's island model (Wright 1943), and to test the statistical significance of Slatkin's (1993) isolation-by-distance (IBD; geographical distance was estimated by following major coastline contours using the ruler tool in Google Earth) among all subpopulations and among collections taken each year.

To further investigate population genetic structures, we used two methods to analyze temporal and geographical variation. First, STRUCTURE (Bayesian method, version 2.3.2 (Pritchard et al. 2000, Hubisz et al. 2009)) was used to test for temporal (among-year) differences in subpopulations, combining the data for collections within each year, and in collections within each subpopulation (for subpopulations with multiyear collections) and to test for geographical differences among subpopulations and collections (for both, all years combined), and for collections taken within each year. In STRUCTURE, we ran the Monte Carlo Markov Chain for 20⁵ iterations after a burn-in period of 10⁵ iterations for a range of number of populations (1 - n, where n is the number of collections or subpopulations in the analysis) using the default correlated allele frequencies and admixture options. The most likely number of populations (K) was estimated using the method of Evanno et al. (2005). Second, AMOVAs (ARLEQUIN, version 3.6 (Excoffier & Lischer 2011)) were conducted prior to tests to distinguish statistically different groups of pairwise genetic distance values (both Nei's D and Θ;_ST Unstructured AMOVAs were performed on the same data sets as those used for the STRUCTURE analyses. When an AMOVA revealed significant differences in frequencies among basic elements (years, collections, or subpopulations), we used the K-W test or Wilcoxon's 2-sample test (Sokal & Rohlf 1995), as appropriate, to locate significant differences among mean pairwise genetic distances. We tested for among-year differences in mean genetic distances among all collections taken within each year and among all collections within each year for each multiyear subpopulation (i.e., subpopulation represented by 2, 3, or 4 collections). To reveal geographical patterns, we conducted a hierarchical analysis in which we first tested for significant differences among means of the pairwise genetic distances for collections and, separately, for subpopulations, assuming no population structure. We then tested for higher level population structure among groups of collections or subpopulations as defined by the initial no-structure tests. For all calculations of mean pairwise genetic distances, we set individual negative pairwise genetic distances to 0 and transformed the data for the tests that included 0 values using log (n + 1). Last, we tested for significant departure of each pairwise Θ;_ST value from pairwise Θ;_ST = 0 (ARLEQUIN).

A complication of the second method was that the means calculated for some K-W tests involving pairwise genetic distances were not independent because some pairwise distances were used to calculate more than 1 mean within a particular analysis. However, it was obvious that some collections were sufficiently genetically differentiated from, or similar to, most or all others to drive the pairwise genetic relationships that included those collections, and that a multifaceted pattern of connectivity was present. Therefore, we opted to conduct the analyses because the effects of single populations or population clusters can be lost in general IBD analyses (Bossart & Prowell 1998), particularly when population genetic relationships are not linear geographically, as were ours. We used this approach also because it capitalized on many advantageous features associated with pairwise genetic distances based on statistics related to F_ST (see also Neigel (1997, 2002)), such as Nei's D and Θ;_ST. First, the nonrandom distribution of small genetic differences is more convincing than a single tablewide genetic distance value calculated for an entire data set (Palumbi 2003). Second, underlying properties of dispersal can be revealed because genetic distance is frequently correlated with dispersal ability (Waples 1987, Bohonak 1999) and because even small pairwise distances indicate very little gene flow if they are significant (Palumbi & Warner 2002, Palumbi 2003), particularly if they are corroborated by environmental or biological data (Hellberg 2006, Selkoe et al. 2010, White et al. 2010b). Third, the relative differences among genetic distances speak more about population genetic structure than the absolute magnitude of those values (David et al. 1997, White et al. 2010b). Fourth, F_ST is relatively independent of the mutation process, so different genes in the same species may be compared (Pannell & Charlesworth 2000). Fifth, and last, population fluctuations can increase pairwise genetic distance estimates (McCauley et al. 1995, Pannell 2003), augmenting their value for interpreting population dynamics. The limitations of using pairwise genetic distances are that individual pairwise values may not be reliable; values are based on the assumption that the overall population is in equilibrium (Hellberg 2006); values can be affected by levels of variation within, as well as between, populations (Charlesworth 1998); and values can be inflated by anything that reduces within-population variation (Riginos & Nachman 2001). Thus, individual estimates of pairwise genetic distances may not be exact, but overall patterns that have statistical significance can provide insights into population boundaries.

While conducting the analyses that grouped pairwise genetic distance means, we noticed that means of collection-level pairwise genetic distances and variation around the means were possibly related. To test this idea, we performed correlations on the mean genetic distances versus their SDs using 4-y collection pairs only, to eliminate any potential bias resulting from overall interannual differences in pairwise genetic distances.

Mitochondrial DNA

We calculated haplotype frequencies using REAP (version 4 (McElroy et al. 1992)) and used both Monte Carlo simulation (in REAP) and the R × C test to search for significant differences in haplotype frequencies among collections and subpopulations. Haplotypes other than the common haplotype were combined for the R × C tests. We calculated haplotype diversity (h) and nucleotide diversity (π) for each collection and subpopulation, and calculated pairwise π (Nei 1987) for each collection and subpopulation pair (REAP). The h and π values were tested for correlation with sample size at both collection and subpopulation levels; and 4-y subpopulations were tested for significant differences in mean h and π among years. Both diversity measures were tested for significant outliers within years using Dixon's method (Sokal & Rohlf 1995).

We calculated pairwise genetic distances (Φ_ST (Cockerham 1969, Cockerham 1973)) and nucleotide divergences (d (Nei 1987); REAP) and explored the pattern of mtDNA diversity by testing the Φ_ST values for significance (exact test, ARLEQUIN, version 2.0 (Schneider et al. 2000)) and comparing statistically the means of appropriate combinations of pairwise n values and pairwise d values for significant differences within and between years, collections, and subpopulations using the K-W test, similar to our approach for allozyme-locus pairwise genetic distances. We analyzed IBD by testing the geographical distances of (separately) all, within-year, and 4-y subpopulation pairwise combinations for correlation with the analogous linearized pairwise Φ_ST values (Slatkin 1993), and we tested for increasing genetic distance over time by substituting the number of years between collection pairs as the distance measure (all collection pairs and only 4-y collection pairs; Mantel test in ARLEQUIN). We also conducted AMOVAs (ARLEQUIN) on the same sets of data as those used for the allozyme-locus AMOVAs, but using pairwise Φ_ST values.

External Factors

We mapped the location and extent of the seagrass beds according to Yarbro and Carlson (2013), and formed consensus maps of seasonal and unusual water currents based on the extensive océanographie information available for the West Florida Shelf (see Results, External Factors). For each year, we constructed maps of the tracks of tropical cyclones, obtained from the U.S. National Oceanographic and Atmospheric Administration (NOAA) website ( http://www.nhc.noaa.gov/pastall.shtml) and noted their timing; mapped the distribution and intensity of red tide blooms (drawn from direct cell density counts taken at precise locations during specific days; FWRI, unpubl. data), and documented their timing; and graphed the abundances of bay scallop adults (mean number/600 m²; n usually 20/600-m² transect per collection) and spat (mean number per collector per 42-day deployment period; range, 12–27 collectors per location) using data drawn from FWRI (unpubl. data) and annual reports (Arnold 1994, Arnold et al. 1995, Arnold et al. 1996, Arnold et al. 1997, Arnold et al. 1998b, Arnold et al. 1999).

Amounts of monthly rainfall were compiled from records for Florida maintained by NOAA. The U.S. National Oceanographic and Atmospheric Administration partitions Florida into 7 geographical rainfall divisions. To categorize rainfall amounts, for each division we calculated the average precipitation (±SD) separately for each month using the monthly amounts from 1994 through 1998. Then, separately for each division-month combination, we classified rainfall amounts as follows: very wet, ≥(average + SD) × 1.25; wet, ≥(average + SD) and <[(average + SD) × 1.25]; average-wet, ≥average and <(average + SD); average-dry, ≥(average - SD) and <average; dry, ≥[(average -SD) × 0.75] and <(average - SD); very dry, <[(average - SD) × 0.75]. Because long-term average rainfall varies among months and divisions, a specific rainfall amount may be classified as, for example, dry for a given month/year/division but average-wet for a different month/year/division.

Genetic Variation

Allozyme-locus allele frequencies for the 10 polymorphic loci are given for bay scallop collections and subpopulations in Table 2. The frequency of the most common allele was ≥0.99 in two aspartate aminotransferase and two malic dehydrogenase loci, general protein, and in the hexonate dehydrogenase, iso-citrate dehydrogenase, and superoxide dismutase loci. No geographical patterns in levels of polymorphism were discerned. Only 3 alleles were private alleles.

Within subpopulations, numerous collections varied significantly between years in allele frequencies, and five collections varied between years in genotype frequencies. The allelic variation was concentrated in panhandle collections; the geno-typic variation was in panhandle collections only (both principally SJ). Most allelic and genotypic variation was at specific loci (e.g., LAP). All polymorphic loci except PGD differed significantly in allele frequencies among subpopulations.

Numerous loci did not conform to H-W equilibrium (Table 2), all but one resulting from heterozygote deficits. The percentage of loci deviating from H-W equilibrium in 1998 collections was significantly lower than the percentages in collections from 1996 or 1997 (Table 3A). Collections from Steinhatchee and St. Joseph had the greatest percentages of loci out of H-W equilibrium (Table 3B). Overall, the mean percentage of loci out of H-W equilibrium for 3-y- and 4-y subpopulations (70%; SD, 9%) was significantly greater than for 1-y and 2-y subpopulations (28%; SD, 12%; t-test for small sample sizes, P < 0.001 (Sokal & Rohlf 1995)).

Few measures of genetic variability for the Florida Gulf collections or subpopulations were related to sample size, geography, or time. At the collection level, correlated with sample size (P < 0.0001, r = 0.73). At the subpopulation level, correlated with sample size, year (both positively), and latitude (negatively) (P range, 0.05–0.001). The latitudinal and temporal correlations reflected the relatively high level of allozyme allelic variability in southwestern Florida 1998 collections (Table 2). Neither P₉₅ nor P₉₉ were correlated with any variable at either the collection level or the subpopulation level.

Inbreeding levels varied significantly among years (P = 0.025), collections (P < 0.0001), and subpopulations (P < 0.001). The level of inbreeding was significantly greater in collections from 1996 (F_IS = 0.28) than in collections from 1998 (F_IS = 0.18). Of all collections, PI96 had, by far, the greatest F_IS value (0.40). Inbreeding levels also differed significantly between many collection pairs at very high probability levels.

The BOTTLENECK model assumes that populations are in H-W equilibrium. Because heterozygote deficits are common in our data, and other assumptions of the BOTTLENECK model may have been violated (e.g., Florida Gulf bay scallops show evidence of a recent population expansion (Bert et al. 2011, Figure 3)), results of the BOTTLENECK test were aberrant (e.g., T₂ was significant for many collections, but because of heterozygote deficits and not heterozygote excesses, as the model predicts). Therefore, we relied on other characteristics typically associated with bottlenecked populations to denote bottlenecking: comparatively low levels of polymorphism and numbers of alleles (Nei et al. 1975, Allendorf et al. 2008), and unexpectedly high pairwise genetic distances (Hellberg 2006). One collection, SJ96, had the requisite heterozygote excess (Wilcoxon's test, P = 0.02); PI96 and HE98 exhibited low P₉₉ and ; SJ98 had low P₉₅, P₉₉, and ; and, although not significantly different than H_o values of other 1998 collections, H_o was notably high in FB98. In FB98, less common alleles often had greater frequencies than in other collections (e.g., AAP-2_D, PGM_F), and multiple alleles rare in other collections were missing (Table 2).

The bay scallop samples yielded 43 mtDNA haplotypes (H). The estimated restriction fragment compositions for those haplotypes are presented in Bert et al. (2011, Table 7). Haplo-type frequencies are listed here in Table 4.

Collection and subpopulation h and π values are provided in Table 5. Neither h nor π correlated with sample size at either the collection level or the subpopulation level, nor did mean h or mean π differ significantly among years for the 4-y subpopulations. Mitochondrial DNA diversity was notably low in 1995 collections, greater in ST than in other collections in 1995 (significantly) and 1996, comparatively high in HO and TB (significant for π) in 1997, and greater in SJ and PI (significantly for h) but significantly low (both h and π) in FB in 1998.

Population Genetic Structure

Allozyme Loci

Detecting Structure. Collectively over all loci, allele frequencies differed significantly between 89% of the subpopulation pairs (Table 6). At the tablewide level, TB, SS, and FB differed significantly from all or nearly all other subpopulations, whereas HE differed significantly from SS and FB only. Centrally located subpopulations were notably more similar to each other than to other subpopulations, as were panhandle subpopulations.

Although the estimated overall Θ;_ST value for all subpopulations was small (0.015), at the collection level, pairwise Nei's D values varied more than 30-fold (range, 0.000–0.032) and pairwise Θ;_ST values varied more than 70-fold (range, 0.000–0.071; Table 7A). The overall gene flow estimate was high (effectively, 6 individuals per generation), in part because only polymorphic loci were used (Smith et al. 1997); but gene flow was reduced significantly with increasing geographical distance (y = -0.43x + 2.46, r² = 0.12, P = 0.006). However, the proportion of the relationship explained was small (12%), and the only year in which gene flow and geographical distance were inversely related was 1997 (y = -0.25x + 3.05, r² = 0.21, P < 0.0001).

TABLE 2.

Allele frequencies for allozyme loci highly polymorphic in Florida Gulf bay scallops and associated measures of genetic variability.

continued

TABLE 3.

Temporal and spatial differences in percentage of allozyme loci deviating from Hardy-Weinberg genotype frequency equilibrium expectations.

No STRUCTURE analysis divided the collections temporally or spatially into more than one population. STRUCTURE was designed to sort individuals into populations that conform to H-W equilibrium and to exhibit minimum linkage disequilibrium. The many departures from H-W equilibrium (Table 2) and close genetic relationships (small pairwise genetic distances) of Florida Gulf bay scallop subpopulations likely reduced STRUCTURE'S statistical power (Kalinowski (2011) and references therein). Analyses of molecular variance indicated that variation attributable to differences among subpopulations was significant in 1998 (P = 0.000, 1.75% of all variation) and nearly significant overall (P = 0.056, 0.22% of all variation). These results and the nonrandom geographical patterns of variation in homogeneity of pairwise allele frequencies (Table 6) led us to speculate that significant but complex temporal and spatial population genetic structure existed in Florida Gulf bay scallops. As described in Materials and Methods, and shown in Tables 7 and 8, we used allozyme-locus pairwise genetic distances extensively to define this structure.

Building the Structure. The allozyme-locus pairwise genetic distances varied notably among collections from different years and locations, and between potential parent-offspring pairs (Table 7A). The 1998 bay scallop cohort had a markedly different population genetic structure than cohorts of other years in that many pairwise genetic distances were relatively large. The sequential-year 1997/1998 pairwise Θ;_ST value was significant (P = 0.003), and the mean pairwise genetic distances for pairs that included 1998 collections were significantly greater than the means for pairs with collections from other years (Table 8A1). The high mean collection-level pairwise genetic distances for 1998 were a result of significantly high means for collection pairs from specific subpopulations (SA, SJ, HE; Table 8 A2, analyses 5, 6, 9, 10, 15, and 16). Many pairwise genetic distances that included 1998 collections from those subpopulations were among the highest (Table 7A). In contrast, 1995 was a year when all collections were genetically similar; the mean pairwise genetic distances of 1995 collections were low (Table 8A1).

Collections from ST and HO exhibited greater temporal and spatial genetic homogeneity than collections from other locations. The mean pairwise genetic distances with collections from ST to HO as pair members were low and genetically homogeneous among years (Table 8A2, analyses 11–14), whereas the means for collections from other subpopulations were not (Table 8A2, analyses 5–10 and 15–20). The average genetic distance of ST, HO, and AN parent-offspring pairs was significantly less than the average of pairs from other subpopulations (Table 7A; Wilcoxon's 2-sample test, P < 0.05 for both Nei's D and Θ;_ST), and variation around the within-subpopulation ST- and HO-pair means was less than variation around the means with other pairs (Fig. 2). Most means of pairwise genetic distances that included a collection from either ST or HO were in the lower half of the ranges of means for both genetic distance measures (Table 8B1, analyses 21–24, bold print). At the subpopulation level, the mean pairwise genetic distances of pairs with ST and HO were among the lowest (Table 8B2, analyses 25 and 26, bold print)

The genetic stability of ST and HO through time, and their relatively close genetic relationship to each other and other subpopulations, led us to conclude that ST and HO formed a centralized source population (which we named Core; Table 7A). We then focused on defining the relationship between Core and other subpopulations. We reasoned that, if a subpopulation was part of Core, the mean pairwise genetic distance that included pairs with the collections from that subpopulation would not differ significantly from the means that included the collections from ST and HO; the mean distances would be similarly low. We first considered the remaining three 4-y sub-populations (Table 8C1, analyses 27 and 28). Nei's D indicated that the means of the pairwise genetic distances with collections from SA or SJ as pair members differed significantly from the mean of pairs with HO collections as pair members. Other previous analyses also demonstrated differences between SA and SJ versus ST and HO (e.g., Fig. 2, Tables 6 and 7A). Collectively, the evidence showed that SA and SJ were not members of Core. We therefore designated SA and SJ as a northern peripheral population and named it Panhandle.

TABLE 4.

Haplotype frequencies for Florida Gulf bay scallops based on RFLP analysis.

continued

To evaluate PI, we compared the mean pairwise genetic distance of collections including that subpopulation to the collection-level means of the combined Core subpopulations and of the combined Panhandle subpopulations (Table 8C1, analyses 29 and 30). Assignment of PI remained ambiguous. However, when the mean of the PI pairs was analyzed separately with the means of Core pairs and Panhandle pairs, the mean of PI pairs grouped with the mean for Panhandle pairs and not with the mean for Core pairs (Table 8C1, analyses 31–34). Based on this result and on PI generally higher collection-level pairwise genetic distances (Table 7A), we designated PI as a southern peripheral population and named it Southwest. Because the mean genetic distances of pairs with PI collections were nearly identical to the means of pairs with Panhandle collections, we grouped collection pairs with SA, SJ, or PI as pair members (collectively, peripheral collections) to categorize the remaining subpopulations.

The mean pairwise genetic distances of CI-collection/Core-collection pairs and CI-collection/peripheral-collection pairs grouped with the mean for peripheral collections and separately from the mean for Core collections (Table 8C1, analyses 35 and 36). That, and CI location within the same bay system as SA (Fig. 1), verified that CI was a component of Panhandle.

Because the 2 HE collections differed remarkably in their pairwise genetic distances with other collections (Tables 7A and 8B1, analyses 21 and 22), results for the comparisons of the HE-collection/Core-collection means or HE-collection/peripheral-collection means versus the means for Core collections or peripheral collections were ambiguous (Table 8C1, analyses 37–42). However, analyses equivalent to those conducted for PI pairs demonstrated that AN—HE's neighbor to the south (Fig. 1)—was affiliated with Core subpopulations and not with peripheral subpopulations (together, SA, SJ, and PI; Table 8C1, analyses 43–48). In addition, the means and variances of AN—AN collection pairs (D, 0.003; Θ;_ST, 0.004) were nearly as low as those of Core-population ST-ST pairs (Fig. 2); the means of AN collections paired with other collections were as low as those of Core collections paired with other collections (Table 8B1, analyses 21 and 22); and AN allele frequencies were generally similar to those of Core subpopulations (Table 6). Therefore, we included both HE and AN in Core.

Group membership tests performed for 1-y subpopulations (i.e., those represented by single collections) CK, TB, and SS were inconclusive. Those subpopulations were assigned to populations based on their geographical location and other analyses. Cedar Key bay scallops reside deep within Core, and nearly all CK/Core-member pairwise genetic distances were at or near 0 (Table 7A); thus, CK was included in Core. Tampa Bay and SS are outside the coastal region likely to receive recruits regularly from Core but are somewhat more likely to receive recruits from PI (as a result of regional currents, explained later); and both are extinct intermittently. Therefore, they belong with PI in Southwest. In contrast, the 1-y FB subpopulation clearly was genetically distinct from both Core and peripheral collections. Many pairwise Θ;_ST values with FB as a pair member were significant (Table 7A, B); mean genetic distances of FB pairs were significantly high (Table 8B, analyses 21, 22, 25, and 26); and mean genetic distances of FB paired with either Core or peripheral collections were uniquely greater than the means of either Core or peripheral collection pairs (Table 8C1, analyses 49–50).

TABLE 5.

Mitochondrial DNA diversity estimators for Florida Gulf bay scallop collections and subpopulations (all years combined) based on RFLP analysis.

Comparing mean collection- and subpopulation-level pair-wise genetic distances among the 4 populations (Panhandle, SA-SJ; Core, ST-AN; Southwest, TB-PI; and unto itself, FB) exposed population-level relationships. The means of pairwise genetic distances between Core collections were significantly less than the means of other collections from other populations (Table 8C2, analyses 51–54), and the means of collection and subpopulation pairs with FB as a pair member were significantly greater than the means of other populations (Table 8C2, analyses 51, 52, 55, and 56). At the population level, FB was genetically more similar to peninsular populations, particularly Core, than to Panhandle (Table 8C2, analyses 55 and 56).

Last, pairwise genetic distances decreased through time; 1-or 2-y subpopulations were less genetically homogeneous with each other or with 3- and 4-year subpopulations than 3- or 4-y subpopulations were with each other (compare Tables 7B and 7C).

Mitochondrial DNA

The relatively high pairwise nucleotide diversity values for pairs with Core collections or peripheral collections in 1998 (Table 9A) reflected their individual relatively high π values (Table 5). Similarly, the low π value of FB98 resulted in low pairwise π values for FB98 pairs. Excluding pairs with FB98 (which were all <1 %), the mean π value for pairs from 1998 was significantly greater than the mean for pairs from other years (all collections, P < 0.001; 4-y collections only, P < 0.01), principally because of significantly high means of pairs with SJ98 (P < 0.0001) and PI98 (P < 0.02). At the subpopulation level (Table 9B), pairs with the single-year SA and TB had the highest π values, but only the mean of pairs with FB as a member differed significantly from the means with other subpopulations as pair members (FB pairs significantly low; P < 0.001). Among the 4-y subpopulations, within-subpopulation mean pairwise π values were significantly greater for ST collection pairs than for PI collection pairs (P = 0.05). Other than 1998, Core sub-populations generally had greater mtDNA diversity than peripheral subpopulations.

Thirteen of the 17 significant collection-level Φ_ST values had FB98 or SJ98 as a pair member (Table 9A). All PI-FB pairs were significant, and the distance between parental SJ95 and potential offspring SJ96 was highly significant. Within years, pairwise distances were significant for 1998 collections only (SA98/SJ98, SJ98/AN98, PI98/FB98). At the subpopulation level (Table 9B), 15 of the 17 positive Φ_ST values and the three significant values involved peripheral subpopulations. Few tests for significant differences among mean pairwise Φ_ST values were significant, and the pattern of significance was not clear (results not shown). Nonzero potential parent-offspring Φ_ST values occurred in all multiyear subpopulations except PI (Table 9A); some values were significantly high (SJ95/SJ96, ST96/ST97).

Nucleotide divergence was low. Most pairwise, collection-level d values were 0 or slightly negative (mean of all d, 0.0002; SD, 0.0002). The mean d for pairs with FB was 5 times to 10 times greater than the means of other pairs (data not shown). Of the 35 positive pairwise d values, 54% had FB as a pair member (mean d, 0.001; SD, 0.0005) and 34% had an ST collection (mean d, 0.0002; SD, 0.0002) as 1 or both pair members. The low nucleotide diversity of FB and high nucleotide diversity of ST contributed to the high frequencies of pairs with nonzero d values that included those collections.

Linearized pairwise Φ_ST values and geographical distances were related significantly for the 1997 collections (r = -0.46, P = 0.02) and 1998 collections (r = -0.28, P = 0.05). Geographical distance explained 21 % of the variation in genetic distance in 1997 but only 8% in 1998. No tests for temporal IBD were significant. No AMOVA results were significant; variation within populations accounted for nearly 100% of the variation in all analyses.

TABLE 6.

Significance levels for subpopulation-level pairwise tests for homogeneity of bay scallop allozyme-locus allele frequencies.

Interannual Variation

The allozyme-locus pairwise genetic distances revealed notable interannual variation in genetic connectivity between collections and in the structure of that connectivity (Fig. 3, pairwise genetic distance matrices). In 1995, values for all pairwise genetic distances were small (Fig. 3A); and no pairwise Θ;_ST values differed significantly from 0. In both 1996 and 1997, most collections were similar genetically, but collections from both extremes of the sampling range were more genetically differentiated from other collections, and the pattern of differentiation was reversed between years (Fig. 3B, C). In 1996, collections from the two northernmost adjacent subpopulations were significantly differentiated, and the collection from the southernmost subpopulation was differentiated from all collections north of HO. In 1997, collections from the two adjacent southernmost subpopulations were significantly differentiated, and the collection from the northernmost subpopulation was well differentiated from collections taken south of ST. The 1998 matrix of pairwise genetic distances (Fig. 3D) illustrates the patterns of significance that differentiated 1998 collections from others in Tables 8A and 8B (italicized). Pairwise genetic distances were, on average, much greater than in other years because collections from Panhandle were significantly differentiated from each other and from most other collections; and collections from some geographically proximal subpopulations were highly differentiated (notably, SA98/SJ98, HE98/AN98, SS98/FB98).

The geographical extent of the most genetically similar collection pairs (i.e., those considered to constitute Core) varied among years (Fig. 3A–D, boxes enclosing collections in pair-wise genetic distance tables and dashed lines enclosing hatched areas on Florida maps). In 1995, Core was discernible although all pairwise genetic distances were small. In 1996, Core extended from Panhandle (SJ) to AN, its broadest geographical extent. In 1997, Core was again located within the Big Bend but AN was excluded. In 1998, Core was small geographically (ST-HO) as a result of high levels of differentiation between HE98 and AN98 and other Core collections.

To gain a perspective on regional gene flow each year, we averaged the pairwise Θ;_ST values within and between the defined populations. Genetic distances are inversely related to gene flow (Slatkin 1985) so relatively low genetic distance values connote relatively high gene flow (Fig. 3A–D, Florida maps: within-population gene flow represented by colored boxes around subpopulation abbreviations; between-population gene flow represented by colored lines). Gene flow was always high within Core, ranged from high (1995) to very low (1998) within Panhandle, and was moderate within Southwest. Gene flow between Core and Panhandle ranged from high (1996) to low (1998), whereas gene flow between Core and Southwest was more consistent among years (moderate, but high in 1997). Gene flow was high between Panhandle and Southwest only when it was high throughout the study range (1995); when populations were structured, gene flow between Panhandle and Southwest tended to be low. The high gene flow between FB and Core (Fig. 3D) may have been an artifact of high frequencies of common alleles in FB, or FB may have received recruits from Core. Between FB and peripheral populations, including geographically adjacent Southwest, little genetic exchange occurred.

External Factors

Expansive offshore seagrass flats extend between the St. Marks River and Anclote Estuary (Fig. 1). Both westward in panhandle waters and southward to waters off Cape Romano, substantial seagrass beds are within bays. From Cape Romano to Cape Sable, seagrass beds are very sparse; the seagrasses that exist are thinly distributed and usually are not Thalassia or Syringodium. In Florida Bay, only the westernmost and some southernmost basins typically contain dense seagrass beds; but, westward of Florida Bay, extensive seagrass beds exist along the Gulf side of the Florida Keys and throughout the Keys.

In the areas inhabited by bay scallops, seasonal currents interact with shoreline contours to form 4 regional current domains (Fig. 4A–D): (1) from CSB westward (includes SA-SJ), (2) the Big Bend (includes ST-HO and sometimes HE and AN); (2) southwestern Florida (includes TB-PI and sometimes HE and AN); and (4) Florida Bay (includes FB). From CSB westward, water currents are almost always strong, usually running westward from late spring through fall and eastward during winter and early spring (He & Weisberg 2002, He & Weisberg 2003). Overall, currents in this region run westward more than eastward (Ohlmann & Niiler 2005). When currents run eastward, some water moves around into the Big Bend, but a significant component veers offshore at CSB in an offshore jet (Fig. 5A), which generates upwelling off CSB (Ohlmann & Niiler 2005). From spring through fall or early winter (Yang & Weisberg 1999, He & Weisberg 2003), slow, variable currents moving generally northwestward to northeastward push water into the north end of the Big Bend, where it piles up and generates downwelling (Li & Weisberg 1999) (Fig. 4A–C).

TABLE 7.

Matrices of Florida Gulf bay scallop allozyme-locus pairwise genetic distance estimators (Θ;_ST, above diagonal; Nei's D, below diagonal).

TABLE 8.

Statistically homogeneous groupings of allozyme-locus average pairwise Nei's D (D) and Θ;_ST values (from Table 7A, B) for bay scallop collections and subpopulations grouped by year, geographical location, or previous grouping (see Results).

continued

continued

continued

Figure 2.

Relationship between the mean and variation around the mean in pairwise collection-level genetic distance measures of bay scallop subpopulations sampled 4 y (from Table 7A). Labeled points are the means of all within-subpopulation collection pairs (e.g., SA/SA is the mean of values for SA95/SA96, SA95/SA97 … SA97/SA98). Other points are the means of other analogous groups of collection pairs (e.g., means of SA95/SJ95 … SA98/SJ98). Larger points are proportional to the number of pairs with the same means.

From fall or early winter to spring, prevailing northerly winds drive an anticyclonal (clockwise) gyre—which is strongest in October through December (Yang & Weisberg 1999) and is enhanced by tides (He & Weisberg 2002)—that moves southeastward alongshore to the vicinity of Homosassa Bay or Anclote Estuary (Yang & Weisberg 1999), where it veers offshore and moves northwestward until encountering CSB, which deflects the water eastward to complete the gyre (Yang et al. 1999) (Fig. 4D). The gyre can generate upwelling off the Big Bend coast (Yang et al. 1999, Weisberg et al. 2001, He & Weisberg 2003); but, off CSB to Deadman Bay, currents typically remain weak to near 0, particularly during transition months (Hetland et al. 1999, Li & Weisberg 1999, Yang & Weisberg 1999, Weisberg & He 2003) (Fig. 4A, C), and annual net flow is usually near 0 (Ohlmann & Niiler 2005, Liu & Weisberg 2007). Currents between Deadman Bay and the Homosassa-Bay/Hernando vicinity are also very low if the wind is directly onshore or offshore (Li & Weisberg 1999) (Fig. 5B, C).

Summer and winter currents are generally stronger from Tampa Bay southward than in the Big Bend (Yang & Weisberg 1999), and both velocity and tidal amplitude increase with movement southward (He & Weisberg 2002) (Fig. 4B, D). A strong, nearshore, southward-moving, upwelling-favorable coastal jet (Yang et al. 1999) is a common feature during fall, winter (strongest), and spring, particularly in waters between Tampa Bay and Pine Island Sound or Cape Romano (Li & Weisberg 1999, Liu & Weisberg 2005, Weisberg et al. 2005, Weisberg et al. 2009b, Liu & Weisberg 2012) (Fig. 4A, C, D). The coastal jet, together with the Big Bend gyre, generates upwelling off Tampa Bay (strongest October through January). During summer, the coastal jet is weaker, farther offshore, downwelling favorable, and northward flowing (Weisberg et al. 2009b, Liu & Weisberg 2012) (Fig. 4B). Net water movement is south to southwest offshore (He & Weisberg 2002, Ohlmann & Niiler 2005), but very nearshore, northward or shoreward countercurrents can develop (He & Weisberg 2002, Liu & Weisberg 2005).

Current flow in the western basins of Florida Bay is commonly westerly (Lee & Smith 2002), from the basins to open water or south-to-southwesterly through channels separating the Florida Keys (Lee & Smith 2002, Zheng & Weisberg 2012) (Fig. 4A–C). Inflow into western basins occurs sporadically during winter and spring, when cold front-associated westto-northwest winds blow (Fig. 4D). Transport off the Ten Thousand Islands and Cape Sable is offshore (fall), northward (summer), or southwestward (winter/spring), bypassing Florida Bay (Lee & Smith 2002).

Tropical cyclones, red tides, and floods produced by heavy rains could affect a bay scallop cohort's population genetic structure by reducing abundance of, or eliminating, its parental generation as adults or the cohort generation itself at some life stage. Tropical cyclones also generate strong or atypical currents that can transport a cohort's larvae to unusual places, particularly those occurring during October. With this timeline in mind, and considering that tropical cyclone activity almost always occurred after we sampled in June, we included in the same figure as a cohort's pairwise genetic distances the tropical cyclone, red tide, and rainfall activity from the previous year (Fig. 3A–D). Occasionally, considering the timing of events that could affect a cohort's population genetic structure required viewing figures from two consecutive years.

Tropical cyclone activity varied among years (Fig. 3A– D, inset maps and black lines on main maps). Tropical cyclones swept through panhandle waters every year but through peninsula waters only some years. Throughout summer and fall 1995, the Gulf was besieged by tropical cyclones. During 1997, Florida was nearly unaffected by tropical storms.

The extent and intensity of red tide outbreaks varied greatly among years (Fig. 3A–D; main maps, enclosed areas shaded by diagonal lines and accompanying notes stating months of Karenia brevis outbreaks). Blooms occurred in panhandle waters most years, but location and intensity varied. The Big Bend was seldom affected by red tides. Between Anclote Estuary and Pine Island Sound, red tide blooms were broadly distributed and prolonged every year, particularly from Tampa Bay southward. Blooms could occur at any time of the year, but most occurred September through December. Blooms usually followed coastal contours, running alongshore and extending out into the ocean for varying distances (Fig. 3A–C). In 1997, red tide blooms assumed a very different configuration, extending from nearshore to offshore, with little alongshore development (Fig. 3D).

TABLE 9.

Matrices of ranges of Florida Gulf bay scallop pairwise nucleotide diversity (π, above diagonal) and genetic distance (Φ^ST, below diagonal, ×10³) based on mtDNA RFLP analyses.

Of the 7 NOAA rainfall divisions, those affecting bay scallop habitats are divisions 1–5 (Fig. 3). Monthly rainfall patterns fluctuated greatly among those divisions within and between years (Fig. 6). In general, more rain fell May through August than during other months; but, in some years, one or more months between September and January were very rainy. Of note, heavy rains occurred in the panhandle and Big Bend area during summer and in southwest Florida during fall and early winter in 1994, in the panhandle during spring and in west-central and southwest Florida from summer through fall 1995, in the Big Bend area during March and from west-central through southwest Florida during early summer 1996, and from the panhandle through southwest Florida during April and early winter 1997 (Fig. 3A–D, rainfall notes on main maps; Fig. 6).

The maximum abundance of adult bay scallops varied nearly 5-fold among years, and abundance varied even more among subpopulations within some years (Fig. 7). Preharvest adult abundance was high in a Panhandle subpopulation and/or ST 1994 to 1996, but was low in all subpopulations in 1997 and, particularly, 1998. Postharvest abundance was very low compared with preharvest abundance in 1994 and 1995, declining as much as 95%, but was comparable with preharvest abundance in 1996 and 1997. Spat abundance also varied greatly among years and could be exceptionally high in single subpopulations during specific settlement seasons (e.g., SJ95–96, TB97–98).

Population Genetic Structure and Environmental Influences

The Metapopulation

Interlocational, interannual, and intergenerational variation in population genetic structure is common in shallow-water marine organisms (e.g., Kordos & Burton 1993, Edmands et al. 1996, Moberg & Burton 2000, Carlin et al. 2003, Galindo et al. 2010); and genetic differentiation, often structured as a metapopulation, occurs in the face of potential gene flow in various scallop species (e.g., Heipel et al. 1998, Ridgway & Dahle 2000, Orensanz et al. 2006, Marko & Barr 2007, Owen & Rawson 2013). Florida Gulf bay scallops form a complex, mixed-model metapopulation (Harrison & Taylor 1997)—a combination between a pure source-sink metapopulation, in which all gene flow is from a donor source population to recipient sink populations (Tero et al. 2003), and a classic metapopulation, in which all populations have roughly equivalent potentials to contribute recruits to all other populations (Levins 1970).

Figure 3.

Interannual variation in cohort bay scallop allozyme-locus population genetic structure and in external factors potentially contributing to that variation over the time frame relevant to cohort larval dispersal and cohort survival. Matrices: Pairwise genetic distances (×10³; Θ;_ST above diagonal: blue, ≤0–9; green, 10–19; yellow, 20–29; red, 30–71; Nei's D below diagonal: blue, ≤0–4; green, 5–9; yellow, 10–19; red, 20–32) representing population genetic structure. Collections included in the Core population for each year are outlined in each matrix. Populations as defined by all analyses are shown in column and row titles (Pan., panhandle; SW, southwest). Extent of Core in any year may differ from analysis-defined Core. Collection (Coll.) abbreviations are defined in Table 1. (B–D) Boxed Θ;_ST values differ significantly from Θ;_ST = 0. Inset maps: Full regional paths of tropical cyclones; S, tropical storm; H, hurricane; month abbreviation, month of principal activity affecting Florida. Maps of Florida: Colored boxes around collection abbreviations, mean within-population relative genetic distances; colored lines between populations, mean between-population relative genetic distances (Θ;_ST coloration, as described earlier). Dotted area bounded by dashed lines, geographical extent of Core population. Hatched areas bounded by solid lines, maximum extents of red tide blooms; concentration ranges given in (A) apply also to (B–D). Dark gray lines, NOAA rainfall divisions, identified by boxed numbers; months of exceptionally high rainfall displayed within affected rainfall districts (Fig. 6). Black lines, pathways of centers of tropical cyclones.

The Florida Gulf bay scallop metapopulation is also structured hierarchically (per Bilodeau et al. (2005) and Kritzer and Sale (2006)). In such metapopulations, gene flow and genetic isolation are dynamic processes that can occur at different scales (Le Corre et al. 2012), and can vary temporally and spatially in strength and duration. Genetic differentiation can be greater over short distances than over long distances and can be highly variable between cohorts from the same subpopulation. In the Florida Gulf bay scallop metapopulation, Core is the main source population; but, within every population is at least one subpopulation that could act as a local source (ST or HO in Core, SJ in Panhandle, PI in Southwest, and FB as its own source).

Figure 4.

Consensus seasonal current flows along the West Florida Shelf. (A) Spring transition: March, April, or May. (B) Summer: April, May, or June to August or September. Names and geographical features are mentioned in the text. (C) Fall transition: September or October. (D) Winter: October or November to February, March, or April. Arrows approximately represent relative rate and direction of water flow; dots, very little net current flow. Collection abbreviations defined in Table 1.

Luxuriant seagrass beds are essential in shaping the distribution of Argopecten irradians throughout its range, and currents guide the genetic relationships of many scallop populations, including A. irradians (Kraus et al. 1994, Wolff & Mendo 2000, Bogazzi et al. 2005, Tian et al. 2009, Bert et al. 2011). Vagaries in currents during periods of spawning or larval dispersal lend a perpetual element of intergenerational unpredictability to scallop population genetic structures (Tian et al. 2009); but, for Florida Gulf bay scallops, the location and extent of seagrass beds, together with the circulation patterns of the 4 Florida Gulf seasonal current domains, seem to be the basis for their metapopulation structure. Seagrass beds set the stage for the structure; seasonal water currents help maintain it.

We examined the validity of our proposed structure by submitting the population-level pairwise Θ;_ST data to AMOVA. The structure was highly significant (P = 0.004). Significant pairwise Θ;_ST values (ARLEQUIN; P range, 0.03–0.005) also supported this structural model; Panhandle and Southwest were differentiated genetically from each other but not from Core, and FB was differentiated from all other populations. Overall, gene flow was high within Core, higher between Core and Southwest than between Core and Panhandle, and low to very low within and between peripheral populations (Fig. 8).

Juxtaposed on this population genetic structure is increasing genetic homogeneity over an increasing number of generations. Comparing single-year and multiyear pairwise genetic distances matrices revealed that, collectively over generations, subpopulations attained greater genetic homogeneity. Similar declines in genetic patterning over time were found theoretically by Galindo et al. (2010) in population genetics simulations of the Caribbean reef coral Acropora cervicornis, and empirically by Pruett et al. (2005) for Gulf of Mexico red snapper (Lut janus campechanus). In our study, the increase in percentage of allozyme loci out of H-W equilibrium with an increasing number of generations sampled and the high percentage of loci out of H-W equilibrium in two consistent subpopulations (SJ, ST) together with other evidence (Table 10, characteristic 1) indicate that the increase in homogeneity as generations are amassed may be the result of a Wahlund effect—the mixing of recruits from bay scallop aggregations with different allele frequencies. The aggregated distributional pattern of bay scallops during the spawning season (Marelli et al. 1999) is conducive to this type of genetic variation. Regardless of the reason, the increasing homogeneity among subpopulations as collections are amassed must be a result of a general averaging of the heterogeneous genetic diversities among cohorts.

Figure 5.

(A–D) Other current flows along the West Florida Shelf (depth averaged). Current patterns with sustained north-to-northwesterly wind (most common in winter) (A); strong, sustained west wind; names and geographical features mentioned in text (B); strong, sustained east wind (C); and sustained southeasterly winds (most common in summer) (D). Arrows approximately represent relative rates and directions of water flow; dots, very little net current flow. Collection abbreviations defined in Table 1.

Constituent Populations

Core. Genetic evidence indicates that Core is the largest and most stable population, and it is essentially self-sustaining (Table 10, characteristic 2). Core has the greatest genetic diversity (both allozyme and mtDNA) and connectivity. Although the geographical extent of Core varied among years, it always included the genetically diverse subpopulations ST and HO, indicating that gene flow from these subpopulations is important for maintaining the peripheral populations and the less stable Core subpopulations CK and HE, which exhibit high genetic relatedness to ST or HO and thus probably receive recruits from those subpopulations (Hellberg et al. 2002). Evidence of genetic bottlenecking in HE98 (Table 10, characteristic 3) also implies that Core's less stable subpopulations are sometimes composed of offspring from only a few parental bay scallops. The southernmost AN is sufficiently differentiated genetically from other Core subpopulations in some years so as to be excluded from Core. It is the only Core subpopulation located in an estuary behind barrier islands and thus has physical barriers to gene flow that other Core subpopulations do not have, and it is subjected to red tides more frequently than other Core subpopulations.

Core exists and its discrete subpopulations (separated by 30 km or more (Arnold et al. 1995, Arnold et al. 1997)) have high connectivity because extensive seagrass beds are able to grow in the clean, quiet, shallow, open water of the Big Bend (Iverson & Bittaker 1986). The openness of Core provides the opportunity for current-driven larval dispersal within the population and between it and other populations. However, the upwellings generated offshore of the Cape San Blas and Anclote/Tampa Bay-area headlands and the slow, variable nearshore water currents that the headlands help to create in the Big Bend likely inhibit transport from Core both downstream and upstream (e.g., Cowen & Sponaugle 2009). In almost any current pattern, particles such as larvae are transported around within the Big Bend, and only sometimes are eventually transported out (Ohlmann & Niiler 2005). Particularly during the fall/winter bay scallop spawning/recruitment season, northward-moving currents along the coast that stimulate downwelling in the northern part of the Big Bend (Fig. 4C) and the Big Bend winter gyre (Fig. 4D) could retain larvae within the Big Bend, and southward-moving currents that tend to veer offshore (Fig. 4D) could inhibit Core-to-Southwest larval transport along the peninsula. Sustained northwesterly winter winds could generate currents that promote larval transport from Panhandle to Core (Fig. 5A) and contribute to the intermittent genetic similarities between those two populations.

Red tides, which can render subpopulations extinct at any time of the year, were present in and offshore of Anclote Estuary more so than of other Core subpopulations. AN was almost completely eliminated by red tide in 1996 (Arnold et al. 1997). Red tides contribute to either genetic differentiation or genetic homogenization of subpopulations, depending on the sources and effective population sizes (Hedgecock 1994) of larvae reestablishing the decimated subpopulations. Despite the 1996 near-extirpation, the genetic diversity of AN97 was not notably low, and pairwise genetic distances between AN97 and other Core subpopulations were low (Fig. 3C), suggesting that gene flow was high.

Peripheral populations. Compared with Core subpopulations, peripheral subpopulations undergo greater temporal variation in genetic composition and genetic variation and, consequently, generally have greater pairwise genetic distances within and between subpopulations than Core. All peripheral subpopulations are located in bays where, to different degrees, limited exchange with open water impedes successful recruitment among subpopulations, and variation in salinity and temperature increases bay scallop mortality. Red tides and floods, which are more common in areas occupied by peripheral populations than in Big Bend waters (compare A–D in Fig. 3), also decrease abundance by increasing mortality and, thus, contribute further to the abundance fluctuations and intermittent extinctions experienced by these subpopulations. Consequently, all peripheral populations have multiple characteristics of nonequilibrium recruitment and population instability (Table 10, characteristics 1, 3, and 4), as well as enhanced genetic differentiation between subpopulations (Pannell & Charlesworth 2000).

Overall, similarities in allele frequencies and low subpopulation-level pairwise genetic distances (Tables 6 and 7B) suggest that Panhandle subpopulations could constitute a quasi-independent population. The UPGMA cluster analysis of Bert et al. (2011, Fig. 2A) supports that idea. However, relatively high pairwise genetic distances between potential parent-offspring pairs and between collections taken within a single year (Table 10, characteristic 4, Panhandle), together with evidence of drift or bottlenecking (Table 10, characteristic 3, Panhandle), suggest that subpopulations are less stable than Core subpopulations and that self-seeding or exchange of larvae among subpopulations is highly variable. The broadly ranging connectivity levels between Panhandle and Core (from genetically indistinguishable in 1995 to well differentiated in 1998) illustrate further the variable nature of Panhandle compared with Core. However, the relationship of Panhandle subpopulations to Core varies; SJ is more closely related to Core than SA or CI, as exemplified by its inclusion in Core in 1996.

Figure 6.

(A–D) Annual rainfall for years 1994 to 1997, respectively. Rainfall amounts (in inches, as downloaded from  http://www.nhc.noaa.gov/pastall.shtml) for NOAA rainfall divisions shown in Figure 3. Shading for monthly relative rainfall intensity levels, as defined in Methods and Materials: very wet, white print on black background; wet, white print on dark-gray background; average-wet, black print on medium-gray back-ground; average-dry, black print on light-gray background; dry, black print on white background; very dry, bold black print on white background with black outline.

Figure 7.

Florida Gulf bay scallop adult and spat abundances (note variation in scale among years in spat abundances). Abundances (defined in Methods and Materials) were not available for all collection in all years. The 1994 adult abundance is included because the September abundance of that parental-adult year class spawned the 1994/1995 spat that we sampled subsequently as adults (1995 cohort); other samples of parental adults, spat, and year cohorts are similarly related. Adult abundances: collection abbreviations (x-axes) defined in Table 1; June, scallops surveyed before opening of recreational harvest season; September, scallops surveyed after closing of recreational harvest season; ns, no sample. Spat abundances: x-axes, months of sampling (9, September; 11, November; 1, January; and so forth).

Figure 8.

Relative magnitudes of gene flow pathways for Florida Gulf bay scallops, inferred from appropriate pairwise genetic distances (Θ;_ST which is inversely related to gene flow) in Figure 3, averaged over all years. Colored enclosed shapes, gene flow levels among subpopulations within populations (labeled); colored lines, gene flow between populations; blue, high (average Θ;_ST <0.009); green, moderate (average 0.010 < Θ;_ST < 0.019); yellow, low (average 0.020 < Θ;_ST < 0.029); red, very low (average Θ;_ST ≥0.030). Subpopulation abbreviations are defined in Table 1. Bold print, most likely source of subpopulations within each population; abbreviations for ST and HO, most likely main recruitment sources for the metapopulation also boxed.

Several external factors contribute to the variation in Panhandle's population genetic structure. Cape San Blas and associated water currents tend to isolate Panhandle (Fig. 4) from peninsular populations, and SA and SJ are remote from open water and in different bay systems, rendering the likelihood of recruitment from peninsular populations and the regular exchange of recruits within Panhandle typically low. Panhandle is decimated periodically by floods, which can eliminate or impact populations severely in multiple ways (Etherington & Eggleston 2000, Wolff & Mendo 2000, Paerl et al. 2001, Burkholder et al. 2004, Mallin & Corbett 2006), sometimes for prolonged periods (Litaker & Tester 2003). Floods reduce Panhandle abundance, generating genetic bottlenecks and contributing to genetic drift and instability (Table 10, characteristics 3 and 4). Red tides can also decimate Panhandle subpopulations and inhibit larval dispersal. Nevertheless, Panhandle subpopulations, apparently, can attain very high abundances within 1 generation (e.g., Fig. 7, SJ in 1996), a source of variation that could contribute further to high genetic distances or to generate greater genetic connectivity among Panhandle subpopulations and between Panhandle and peninsular subpopulations, depending on the origins and number of recruits.

Substantial numbers of bay scallops occurred historically throughout bays southward of Core to Pine Island Sound (Haddad 1989, Lu & Blake 1997, Marelli & Arnold 2001). Southwestern Florida seagrass beds are best developed near the seaward mouths of bays and channels (Yarbro & Carlson 2013), so gene flow among Southwest subpopulations should be common. However, Southwest subpopulations are not closely related, nor do they seem to form a quasi-independent population. Red tides, which have frequented the waters between Anclote Estuary and Pine Island Sound for decades (Tester & Steidinger 1997, Steidinger et al. 1998, Landsberg et al. 2009, Walsh et al. 2009), and typical offshore-directed and upwelling currents (Fig. 4), may contribute routinely to the lack of genetic cohesion in Southwest by depleting abundances and inhibiting larval transport. The low mean pairwise Φ_ST value for PI/PI pairs (0.000) compared with analogous Φ_ST means (SJ/SJ, 0.019; ST/ST, 0.016; HO/HO, 0.004 (K-W test separated PI/ PI mean and SJ/SJ mean; P < 0.01)) suggests that PI, the only semipermanent subpopulation in Southwest, is mainly selfrecruiting. But, genetic bottlenecking in PI96 (Table 10, characteristic 3) and chronically low PI census numbers (Arnold et al. 1994, Arnold et al. 1995, Arnold et al. 1996, Arnold et al. 1997, Arnold et al. 1998b, Arnold et al. 1999, Greenawalt-Boswell et al. 2007) indicate that recruitment is usually from small numbers of parental individuals. Allee effects may hinder reestablishment of sustainable, large subpopulations in Southwest after severe depletion (Marelli et al. 1999).

Currents offshore of Tampa Bay and Sarasota Bay are particularly variable and can be adverse to larval recruitment into the bays (Fig. 4). When successful recruitment occurs, it may be from multiple sources. The 1997 Tampa Bay cohort has characteristics of a subpopulation generated from multiple groups of larvae with some level of genetic differentiation between them (Table 10, characteristic 1). Or, recruitment may be successful because larvae bypass unfavorable open-water conditions. The large aggregation in Sarasota Bay (SS98) that occurred the year after the geographically proximal TB97 aggregation was found may have been composed of TB97 offspring. Recruitment, at least to spat size and possibly of TB97 offspring, was successful in nearby Tampa Bay in 1997–1998 (Fig. 7). Tampa and Sarasota bays are connected by an intracoastal waterway. Interbay larval transport could bypass open-water red tides and the unfavorable currents.

Genetic evidence presented here (Table 10, characteristic 4) and in Bert et al. (2011) indicates that FB is isolated from other populations and, along with census data, that FB has undergone repeated abundance reductions. The retention of high allozyme genetic variability indicates that the population crashes are short-term (Cornuet & Luikart 1996, Bowen & Grant 1997, Zenger et al. 2003). Bay scallops were sufficiently abundant for collecting scientific samples from western Florida Bay in 1979 and 1980 (T. M. Bert, unpubl. data), 1990 (Marelli et al. 1997a), 1993 (Blake & Graves 1995), and 1998 (current study), and spat were found in 1996 (Fig. 7). However, bay scallops were apparently absent during at least some intervening years (W. S. Arnold, unpubl. data; J. M. Stevely, Florida Sea Grant Program, pers. comm., January 2013). Although the Florida Keys seagrass beds are among the most extensive on earth (Fourqurean et al. 2002), bay scallops are rarely seen outside of Florida Bay basins, indicating that local recruitment is limited to within and among the basins. The close relationship of FB with Core (low 1998 pairwise allozyme-locus genetic distances) may be an artifact of the high genetic diversity of Core and high heterozygosity of FB. The near-absence of mtDNA diversity in our FB collections and high mtDNA diversity level recorded for FB by Blake and Graves (1995) demonstrate that gene flow from other populations into FB does not occur regularly.

TABLE 10.

Genetic evidence for characteristics of the Florida Gulf bay scallop metapopulation

continued

Florida Bay is isolated in part because no seagrass flats capable of supporting bay scallops exist between the Florida Keys and FB's nearest neighbor, PI, which is more than 250 km northward and at the dispersal limit of some models for habitat-specific marine organisms (e.g., Cowen et al. 2006). Nor are currents conducive to larval transport from subpopulations farther north into western Florida Bay (Fig. 4C, D). Some FB population crashes may be associated with die-offs of seagrass beds in the western basins inhabited by bay scallops. Between 1987 and the early 1990s, a massive reduction in seagrass density and abundance occurred in western Florida Bay; some recovery had begun by 1994 (Hall et al. 1999). The bay scallop aggregation we sampled may have established or expanded since that seagrass die-off.

Interannual Variation

The basic metapopulation structure varies intergenerationally from nearly panmictic to highly subdivided. The interannual variation can be related to the external factors we considered—what happens, and when and where it happens.

In the Florida Gulf, tropical cyclones and seasonal storms are most common during fall and winter (Weisberg et al. 2001, Weisberg & He 2003)—the principal bay scallop spawning/ recruitment season. Tropical cyclones, which move vast quantities of water abruptly, mix water columns down to at least 25 m, and greatly alter or reverse normal current flows (Ohlmann & Niiler 2005, Weisberg et al. 2009a), can modify population genetic relationships by generating currents that transport larvae great distances or to unusual places relative to transport by normal currents (Caputi & Brown 1993, Shenker et al. 1993, Etherington & Eggleston 2000, Etherington & Eggleston 2003, Montane & Austin 2005, Briones-Fourzán et al. 2008, Eggleston et al. 2010). Before the 1994/1995 spawning/recruitment season, 2 tropical storms generated heavy rainfall over the panhandle (Fig. 3A), causing floods that, together with harvesting, reduced SJ parental adult abundance by 96% (Arnold et al. 1995). Then, Hurricane Gordon passed across southern Florida. Its winds may have helped to homogenize the 1995 cohort genetically. As it approached southern Florida, southeasterly winds would have set up northwestward-flowing currents that could move larvae from ST to Panhandle (Fig. 5D). As the hurricane traversed Florida, northeasterly winds could have dispersed the red tide temporarily (Burkholder et al. 2004) in southwestern Florida waters (Fig. 3A) and moved larvae from Core to Southwest. Mitochondrial DNA haplotype and nucleotide diversities were high in ST95 compared with other 1995 cohort collections, as might be expected if ST was the source and other subpopulations were recipients of subsets of the ST larval pool.

Huge tropical cyclones slowly traversing the central Gulf generate sustained winds for days (e.g., Weisberg et al. 2009a). Those moving south to north, such as Hurricane Opal in October 1995 (Fig. 3B) set up opposing regional gyres with rotational directions like those in Figure 5C. As they pass over the panhandle, wind circulation would change to that depicted in Figure 5B. During the principal bay scallop spawning season, these gyres could divide bay scallop larvae into northern and southern groups separated somewhere in west-central Florida waters, a model that fits the genetic distance pattern of the 1996 bay scallop cohort (Fig. 3B). Hurricane Opal may have established two genetically differentiated larval pools that separated ST and Panhandle subpopulations from PI and dispersed the red tide in the waters around Panhandle (red tide was not recorded again after passage of that hurricane;  http://www.nhc.noaa.gov/1995opal.html; FWRI, unpubl. data), allowing Core larvae to enter St. Joseph Bay, genetically homogenizing SJ and Core larvae (as indicated by the exceptionally broad Core observed in 1996). The genetic bottlenecking of PI96, perhaps the result of low parental cohort PI95 abundance (Fig. 7), contributed the genetic differentiation between PI and northerly subpopulations. Similarly, poor recruitment in SA and CI (Fig. 7) may have contributed to differentiation between those subpopulations.

The opposing pattern of genetic relatedness in 1997 compared with that of 1996 may have been influenced similarly by a tropical cyclone, but with a key difference in its path. As Hurricane Opal moved northward toward Florida, the winds affecting nearshore Florida Gulf waters would have been principally southerly, followed by southeasterly, then easterly, as the storm traversed the panhandle. As October 1996 Tropical Storm Josephine moved northeastward toward Florida, winds would have been principally southwesterly (Fig. 5B), followed by northeasterly (Fig. 5C) as the storm crossed northwestern Florida (Fig. 3C). Thus, the gyres generated by Tropical Storm Josephine could have transported larvae from Core to Southwest. Larval dispersal between TB and PI may have been inhibited by the red tide offshore of SS, which developed 3 wk after the cyclone passed ( http://www.nhc.noaa.gov/1996josephin.html; FWRI, unpubl. data).

Winter storms, which have strong northeasterly to north-westerly winds that can be sustained for days and are strongest during October through December (Yang & Weisberg 1999)—the principal spawning/recruitment season—also generate current patterns like those shown in Figure 5A through C. Any of those current patterns would tend to generate gyres that separated larvae into regional groups with population genetic patterns similar to those seen in the 1996 or 1997 cohorts.

By any estimation, the time period from fall 1997 through 1998 was anomalous. Compared with bay scallop collections from other years, 1998 collections had a significantly low overall percentage of loci deviating from H-W equilibrium; significantly high allozyme-locus heterozygosity, particularly in southern subpopulations (HO-FB); significantly low level of inbreeding; significantly high mtDNA nucleotide diversity; and significantly high overall mean allozyme-locus and mtDNA pairwise genetic distances. Many allozyme-locus collectionlevel pairwise genetic distances were among the highest, including some between neighboring Core subpopulations (Fig. 3D), and mtDNA diversities were redistributed (peripheral subpopulations were more diverse). The 1998 cohort exhibited multiple examples of genetic drift, genetic bottlenecking, and low gene flow (Table 10).

Climatically, Florida was affected by a strong El Niño event. In Florida, El Niños alter wind velocities and patterns, change rainfall patterns and amounts, and suppress hurricanes (Schmidt et al. 2001), triggering a cascade of oceanographic and hydrological events that affect dramatically the shallow-water marine environment (Müller-Karger 2000). No tropical cyclones occurred during fall 1997. Thus, seasonal current domains (Fig. 4), which would tend to isolate bay scallop populations, probably prevailed during the spawning/recruitment season. The absence of strong currents also may have allowed red tides to proliferate and assume their unusual nearshore-to-offshore configurations (Landsberg et al. 2009) and to bloom in unusual places. The only year in which red tide bloomed in the Big Bend was 1997. Heavy El Niño-related rainfall caused extensive flooding (Del Castillo et al. 2001, Jolliff et al. 2003), which lowered salinities in bays throughout the Florida Gulf and generated low-salinity plumes that swept along the panhandle west of CSB (Müller-Karger 2000) and down the peninsula from Cedar Key to Anclote Estuary (Jolliff et al. 2003). Nutrient-rich water in flooded panhandle bays stimulated phytoplankton blooms (Gilbes et al. 2002) that produced detritus-laden anoxic bottom layers, which killed benthic-dwelling organisms (Collard et al. 2000), including bay scallops. Freshwater outflows greatly damaged seagrass beds in the Big Bend (Carlson et al. 2003). Prolonged north-to-south current flow generated cold-water upwelling west of CSB and off the peninsula (Weisberg & He 2003). This extraordinary event overload reduced bay scallop abundances (Fig. 7) and apparently curtailed at least some larval dispersal, resulting in the unusual genetic diversity patterns and high levels of genetic differentiation (especially in Panhandle subpopulations), genetic bottlenecking (Panhandle and Core subpopulations), and genetic instability (Panhandle subpopulations) we recorded for the 1998 cohort (Fig. 3D; Table 10, characteristics 3 and 4).

Management of the Metapopulation

Getting the Picture: Long-Term Studies, Multiple Markers, and External Factors Are Important

Identifying and classifying metapopulation structure using population genetics can be difficult in the marine environment because genetic distances between metapopulation components can be small and gene flow estimates high. To provide confidence in identifying and interpreting metapopulation structure of marine organisms, population genetics studies should be broad in geographical scope, extend over multiple years, include multiple genetic markers, incorporate a comprehensive array of statistical analyses, and draw on oceanographic, climatic, ecological, and life history information. The paucity of studies with these criteria belies their utility for management and conservation.

Studies conducted over broad spatial scales can provide a more holistic understanding of recruitment dynamics and can elucidate complicated subpopulation connectivity patterns (Etherington & Eggleston 2000). Population genetic relationships can be fully understood only through multiyear studies (Hellberg 2006, Owen & Rawson 2013), because gene flow and isolation are dynamic processes that vary in intensity and duration through time and space (Pruett et al. 2005), especially when populations are not in mutation-drift equilibrium (Slatkin 1993), as is the case for many shallow-water marine invertebrates (Sponer & Roy 2002). Many population genetics studies have demonstrated the value of using multiple genetic markers to evaluate population genetic structure of marine animals, particularly those using both nuclear DNA and mtDNA markers (e.g., Ward et al. 1994, Edmands et al. 1996, Brookes et al. 1997, Smith et al. 1997, Grant & Bowen 1998, Shaklee & Bentzen 1998, Salini et al. 2006, Bert et al. 2011). In population genetics studies of marine organisms, thorough data analysis is essential because genetic differences are typically very small; any significant degree of genetic differentiation is evidence for restricted gene flow (Palumbi 2003). Even small genetic diversity values can provide insight into the origins, histories, and interactions of populations (Bohonak 1999, Hamm & Burton 2000, Pannell & Charlesworth 2000, Sherman et al. 2008). Moreover, the validity of small differences increases when many different statistics or replicas support a specific interpretation. Last, combining population genetic structure and climatic, ecological, and biological factors may lead to the ability to predict population dynamics for resource management. Using this analytical approach, we were able to define the population genetic structure of Florida Gulf bay scallops, elucidate its annual variation, and document environmental and biological factors of potential influence in shaping the structure annually, represent the Florida Gulf metapopulation gene pool more fully than if we had sampled for a single year, identify the increasing genetic homogeneity among Florida Gulf bay scallop subpopulations over ecological time, and provide recommendations for future research and management of this complex and dynamic metapopulation.